skills.rest
Yoodaddy0311
Yoodaddy0311COMMUNITY
Explore all skills by this author →
GitHub

prompt-caching-strategy

Optimizes prompt cache hit rates by separating static and dynamic instruction content.

3|2|Updated Feb 13, 2026
Search Tags:#context management#claude code#system prompt#claude.md#cost reduction#token optimization#prompt caching
One-click install
npx skills add https://github.com/Yoodaddy0311/artibot --skill prompt-caching-strategy-yoodaddy0311
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: prompt-caching-strategy
Source: https://github.com/Yoodaddy0311/artibot/tree/main/plugins/artibot/skills/prompt-caching-strategy
Command: npx skills add https://github.com/Yoodaddy0311/artibot --skill prompt-caching-strategy-yoodaddy0311

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Large instruction files like CLAUDE.md are re-sent on every turn, inflating token costs and latency when no caching strategy is applied. This Skill provides a concrete layout and budgeting method to maximize cached static content and minimize per-turn recomputation. ## Core Features & Use Cases - Dynamic Boundary Placement: Explains the __SYSTEM_PROMPT_DYNAMIC_BOUNDARY__ marker and how to place stable rules above it and volatile context (git status, dates, memory) below it. - Token Budget Management: Enforces per-file limits (4K chars) and a total budget (12K chars) with a chars/4+1 token estimation formula. - Progressive Skill Loading: Loads only Level 1 skill content (~200 tokens) first, escalating to Level 2 only when needed. - Use Case: A team with a bloated CLAUDE.md restructures it so 70%+ of the system prompt sits in the cacheable static region, cutting prompt costs by roughly 60%. ## Quick Start Ask the AI to analyze your CLAUDE.md and reorganize it so static rules sit above the dynamic boundary for prompt cache optimization.

Frequently Asked Questions about prompt-caching-strategy

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I optimize prompt caching in Claude Code?▼

Place stable instructions like protocols and quality gates above the __SYSTEM_PROMPT_DYNAMIC_BOUNDARY__ marker, and keep volatile data like git status and dates below it. Cached static tokens cost roughly 10x less than uncached tokens.

What is the system prompt dynamic boundary?▼

It is a marker dividing the system prompt into a static region that gets cached across turns and a dynamic region recomputed every turn. Content above the marker must stay byte-identical to preserve the cache.

How large should a CLAUDE.md file be for token efficiency?▼

Keep each instruction file under 4,000 characters and the total across all instruction files under 12,000 characters. Estimate tokens with the formula chars divided by 4 plus 1.

When should I not use prompt cache optimization?▼

Skip it on platforms without prompt caching support, such as non-Anthropic APIs, or for single-turn tasks where the prompt structure changes every request and no stable static prefix exists to cache.

Why does my prompt cache keep missing?▼

Cache misses happen when any dynamic content, such as timestamps or git status, is placed above the boundary, invalidating the entire cached prefix. Move all frequently changing content below the boundary marker.

Related Skills

Explore More Software Engineering Skills→
daymadedaymade
1.3k

claude-md-progressive-disclosurer

Audit CLAUDE.md sections and migrate content to references or skills.

Community
Advanced
fcakyonfcakyon
1.0k

statusline-setup

Configure Claude Code statusline to display model, context, cache, and session cost metrics.

Community
Basic
johnnymo87johnnymo87
4

beads

Persist task context across sessions using Git-backed beads.

Community
Basic
FVossebeldFVossebeld

context-optimization

Reduce token usage by masking, compacting, caching, and partitioning context.

Community
Advanced